移动机器人在未知环境下的同步定位与地图重建方法-一种改进的FastSLAM算法.pdfVIP

第25 卷 第4 期 控 制 与 决 策 2010 年 4 月 Vol. 25 No. 4 Control and Decision Apr. 2010 文章编号: 1001-0920 (2010) 04-0515-06 移动机器人在未知环境下的同步定位与地图重建方法 张 亮, 蒋荣欣, 陈耀武 (浙江大学数字技术及仪器研究所，杭州310027) 摘 要: 在标准FastSLAM 中, 随着重采样次数的增加会出现十分严重的粒子退化现象, 从而导致机器人位姿估计的 一致性很差. 针对FastSLAM算法的这一缺陷, 提出一种改进的FastSLAM算法. 此算法在标准FastSLAM 的重采样条 件判断中, 额外考虑了粒子权重协方差和每个粒子的测量残余一致性, 并且使用指数等级选择算法进行新粒子的生 成. 通过仿真实验可以看出, 改进的FastSLAM算法不但可以明显地提高机器人位姿估计的一致性, 而且能够很好地 保持粒子多样性. 关键词: 同步定位与地图创建；一致性；测量残余一致性；指数等级选择 中图分类号: TP249 文献标识码: A Algorithm for mobile robots’ simultaneous localization and mapping in unknown environments ZHANG Liang, JIANG Rong-xin, CHEN Yao-wu (Institute of Advanced Digital Technologies and Instrumentation, Zhejiang University, Hangzhou 310027, China. Correspondent: ZHANG Liang, E-mail: zliang520@zju.edu.cn) Abstract: In the normal fast simultaneous localization and mapping (FastSLAM) algorithm, the particle degradation phenomena are quite obvious, which leads to the consistency of the robot pose estimation comparatively worse. Therefore, a improved FastSLAM algorithm is proposed, in which the particle weights’ covariance and every particle’s residual consistency are considered to check whether it is the time to do re-sample, and the new particles are produced by using exponential ranking selection method. Simulation results show that the improved FastSLAM algorithm can improve the consistency of the robot pose estimation obviously and keep the diversity of the particle well. Key words: Simultaneous localization and mapping(SLAM) ； Consistency ； Measurement residual consistency ； Exponential ranking selection